Inferring regulatory networks and dynamical activities of transcription factors and miRNAs

Abstract

Transcription factors (TFs) and miRNAs are fundamental regulators of gene expression. Transcription factors are proteins that regulate the process of transcribing genes in the DNA into mRNA, whereas miRNAs mainly regulate genes post transcription by hindering the translation of mRNAs to proteins.

To understand the role and effect that TFs and miRNAs have in regulation of genes, one can map which genes are regulated by different TFs and miRNAs, and then, knowing the regulatory interactions, find a way to evaluate what kind of effect the different TFs and miRNAs have on their target genes in a specific biological context.

In this thesis, we have first worked on the problems of mapping regulatory interactions by predicting transcription factor binding sites in the DNA, and furthering our understanding of what characterises true and functional interactions.

As part of this work, we have developed benchmarks that can rank the performance of algorithms that predict the location of transcription factor binding sites by searching for known sequence motifs, and have sought to explain why different types of algorithms differ in prediction accuracy. We have also benchmarked and improved upon algorithms that predict transcription factor binding sites by locating peak regions in ChIP-sequencing data; and by analysing clustered peak regions, we have pinpointed the chromatin marks that best explain transcription factor association with different regulatory elements in the DNA.

After having compared ChIP-seq data for a set of TFs in different cell types, we have described what characterises cell-type specific ChIP-seq peaks and noise, demonstrated how several different factors contribute to context-dependent regulation, and developed classifiers that can differentiate between peaks that are common to different cell-contexts and peaks that are context-specific. These experiments have improved our understanding on how to map more accurately the location of transcription factor binding sites and what characterises and distinguishes true and functional binding sites from noise.

In the final part of this thesis, we have sought to evaluate the regulatory effect that different TFs and miRNAs have on their target genes by integrating a noise-reduced regulatory network model of both TFs and miRNAs with time-series expression data. In these experiments, we find that our predicted regulatory effects are consistent and correspond well with existing knowledge on the role of specific cell cycle regulating TFs and on the role of global miRNA regulation and its effect on the cell cycle. We also have shown that one role of miRNAs can be to dampen the expression changes caused by transcription factors.

Together, these results contribute to our knowledge on how to map regulatory interactions and the role of transcription factors and miRNAs in regulating the dynamics of gene expression.